改进卷积神经网络模型设计方法
|
摘要
针对现有卷积神经网络模型参数量大、训练耗费时间的问题,提出一种网络串联和并联共用的方法,使用较小的卷积核和较多的非线性激活减少参数量的同时增加网络特征学习能力,提出尺度归一化池化层取代全连接层,避免全连接层参数过多容易导致过拟合的问题,改进后的模型支持训练任意尺寸的图片。实验结果表明,提出方法减少了大量的参数和训练消耗的时间,有效提升了算法的效率。
Aiming at the problems that the existing convolutional neural network model has large parameters and its training is time-consuming,a method of network serial and parallel sharing was proposed,which used a smaller convolution kernel and more nonlinear activation to reduce the parameter quantity and increase the network feature learning ability,and scale normalization pooling layer was proposed to replace the full connection layer,avoiding the problem that the full connection layer parameters are too easy to cause over-fitting,and the improved model supported the training of images of any size.Experimental results show that the proposed method reduces the number of parameters and time spent on training,which effectively improves the efficiency of the algorithm.
Aiming at the problems that the existing convolutional neural network model has large parameters and its training is time-consuming,a method of network serial and parallel sharing was proposed,which used a smaller convolution kernel and more nonlinear activation to reduce the parameter quantity and increase the network feature learning ability,and scale normalization pooling layer was proposed to replace the full connection layer,avoiding the problem that the full connection layer parameters are too easy to cause over-fitting,and the improved model supported the training of images of any size.Experimental results show that the proposed method reduces the number of parameters and time spent on training,which effectively improves the efficiency of the algorithm.
引文
[1]Han X,Zhong Y,Cao L,et al.Pre-trained AlexNet architecture with pyramid pooling and supervision for high spatial resolution remote sensing image scene classification[J].Remote Sensing,2017,9(8):848-870.
[2]Russakovsky O,Deng J,Su H,et al.ImageNet large scale visual recognition challenge[J].International Journal of Computer Vision,2015,115(3):211-252.
[3]Ke H,Chen D,Li X,et al.Towards brain big data classification:Epileptic EEG identification with a lightweight VGGNet on global MIC[J].IEEE Access,2018,6(99):14722-14733.
[4]Szegedy Christian,Liu Wei,Jia Yangqing,et al.Going deeper with convolutions[C]//IEEE Conference on Computer Vision and Pattern Recognition,2015:1-9.
[5]Rezende E,Ruppert G,Carvalho T,et al.Malicious software classification using transfer learning of ResNet-50deep neural network[C]//IEEE International Conference on Machine Learning and Applications,2018:1011-1014.
[6]Jin Y,Kuwashima S,Kurita T.Fast and accurate image super resolution by deep CNN with skip connection and network in network[C]//International Conference on Neural Information Processing,2017:217-225.
[7]He K,Zhang X,Ren S,et al.Spatial pyramid pooling in deep convolutional networks for visual recognition[J].IEEE Transactions on Pattern Analysis&Machine Intelligence,2015,37(9):1904-1916.
[8]Srivastava N,Hinton G,Krizhevsky A,et al.Dropout:Asimple way to prevent neural networks from overfitting[J].Journal of Machine Learning Research,2014,15(1):1929-1958.
[9]Ioffe S,Szegedy C.Batch normalization:Accelerating deep network training by reducing internal covariate shift[C]//Proceedings of the 32nd International Conference on International Conference on Machine Learning,2015:448-456.
[10]Szegedy C,Vanhoucke V,Ioffe S,et al.Rethinking the inception architecture for computer vision[C]//Computer Vision and Pattern Recognition.IEEE,2016:2818-2826.
[11]Filonenko A,Kurnianggoro L,Jo KH.Comparative study of modern convolutional neural networks for smoke detection on image data[C]//International Conference on Human System Interactions.IEEE,2017:64-68.
[12]Cheng G,Han J,Lu X.Remote sensing image scene classification:Benchmark and state of the art[J].Proceedings of the IEEE,2017,105(10):1865-1883.
[2]Russakovsky O,Deng J,Su H,et al.ImageNet large scale visual recognition challenge[J].International Journal of Computer Vision,2015,115(3):211-252.
[3]Ke H,Chen D,Li X,et al.Towards brain big data classification:Epileptic EEG identification with a lightweight VGGNet on global MIC[J].IEEE Access,2018,6(99):14722-14733.
[4]Szegedy Christian,Liu Wei,Jia Yangqing,et al.Going deeper with convolutions[C]//IEEE Conference on Computer Vision and Pattern Recognition,2015:1-9.
[5]Rezende E,Ruppert G,Carvalho T,et al.Malicious software classification using transfer learning of ResNet-50deep neural network[C]//IEEE International Conference on Machine Learning and Applications,2018:1011-1014.
[6]Jin Y,Kuwashima S,Kurita T.Fast and accurate image super resolution by deep CNN with skip connection and network in network[C]//International Conference on Neural Information Processing,2017:217-225.
[7]He K,Zhang X,Ren S,et al.Spatial pyramid pooling in deep convolutional networks for visual recognition[J].IEEE Transactions on Pattern Analysis&Machine Intelligence,2015,37(9):1904-1916.
[8]Srivastava N,Hinton G,Krizhevsky A,et al.Dropout:Asimple way to prevent neural networks from overfitting[J].Journal of Machine Learning Research,2014,15(1):1929-1958.
[9]Ioffe S,Szegedy C.Batch normalization:Accelerating deep network training by reducing internal covariate shift[C]//Proceedings of the 32nd International Conference on International Conference on Machine Learning,2015:448-456.
[10]Szegedy C,Vanhoucke V,Ioffe S,et al.Rethinking the inception architecture for computer vision[C]//Computer Vision and Pattern Recognition.IEEE,2016:2818-2826.
[11]Filonenko A,Kurnianggoro L,Jo KH.Comparative study of modern convolutional neural networks for smoke detection on image data[C]//International Conference on Human System Interactions.IEEE,2017:64-68.
[12]Cheng G,Han J,Lu X.Remote sensing image scene classification:Benchmark and state of the art[J].Proceedings of the IEEE,2017,105(10):1865-1883.